From patchwork Mon Sep 28 05:42:10 2015 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Wengang Wang X-Patchwork-Id: 7275301 Return-Path: X-Original-To: patchwork-linux-rdma@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork1.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.136]) by patchwork1.web.kernel.org (Postfix) with ESMTP id DADF59F36A for ; Mon, 28 Sep 2015 05:40:57 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id B692220386 for ; Mon, 28 Sep 2015 05:40:56 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 61E11203A9 for ; Mon, 28 Sep 2015 05:40:55 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751253AbbI1Fkv (ORCPT ); Mon, 28 Sep 2015 01:40:51 -0400 Received: from userp1040.oracle.com ([156.151.31.81]:29144 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751130AbbI1Fku (ORCPT ); Mon, 28 Sep 2015 01:40:50 -0400 Received: from userv0022.oracle.com (userv0022.oracle.com [156.151.31.74]) by userp1040.oracle.com (Sentrion-MTA-4.3.2/Sentrion-MTA-4.3.2) with ESMTP id t8S5enPF030335 (version=TLSv1 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Mon, 28 Sep 2015 05:40:50 GMT Received: from userv0121.oracle.com (userv0121.oracle.com [156.151.31.72]) by userv0022.oracle.com (8.13.8/8.13.8) with ESMTP id t8S5enSh007397 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=FAIL) for ; Mon, 28 Sep 2015 05:40:49 GMT Received: from abhmp0008.oracle.com (abhmp0008.oracle.com [141.146.116.14]) by userv0121.oracle.com (8.13.8/8.13.8) with ESMTP id t8S5enEv010618 for ; Mon, 28 Sep 2015 05:40:49 GMT Received: from oracle.com (/10.182.64.160) by default (Oracle Beehive Gateway v4.0) with ESMTP ; Sun, 27 Sep 2015 22:40:49 -0700 From: Wengang Wang To: linux-rdma@vger.kernel.org Cc: wen.gang.wang@oracle.com Subject: [PATCH] IPoIB: serialize changing on tx_outstanding Date: Mon, 28 Sep 2015 13:42:10 +0800 Message-Id: <1443418930-18677-1-git-send-email-wen.gang.wang@oracle.com> X-Mailer: git-send-email 2.1.0 X-Source-IP: userv0022.oracle.com [156.151.31.74] Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org X-Spam-Status: No, score=-6.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_HI, T_RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=unavailable version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP The changing on tx_outstanding should be protected by spinlock or to be atomic operations. Such log is found in dmesg: Sep 16 14:20:53 naep11x06 kernel: ib0: queue stopped 1, tx_head 1034733, tx_tail 1034733, tx_outstanding 359 ipoib_sendq_size: 512 Sep 16 14:21:33 naep11x06 kernel: ib0: transmit timeout: latency 9560 msecs Sep 16 14:21:33 naep11x06 kernel: ib0: queue stopped 1, tx_head 1034854, tx_tail 1034854, tx_outstanding 511 ipoib_sendq_size: 512 Sep 16 14:21:38 naep11x06 kernel: ib0: transmit timeout: latency 14568 msecs Sep 16 14:21:38 naep11x06 kernel: ib0: queue stopped 1, tx_head 1034854, tx_tail 1034854, tx_outstanding 511 ipoib_sendq_size: 512 And the send queue of ib0 kept full. When transmit timeout is reported, queue is reported as "stopped", but the IPoIB stuff tx_head and tx_tail points to same value. I am not able to see such numbers in ipoib_cm_tx (for CM) because I have no vmcore. Though I am not quite sure it's caused by parallel access of tx_outstanding(send path VS interrup path), we really need to serialize the changeing on tx_outstanding. This patch also make sure the increase of tx_outstanding prior to the calling of post_send to avoid the possible decreasing before increasing in case the running of increasing is scheduled later than the interrupt handler. Signed-off-by: Wengang Wang --- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 40 +++++++++++++++++++++++---------- drivers/infiniband/ulp/ipoib/ipoib_ib.c | 24 ++++++++++++++++++-- 2 files changed, 50 insertions(+), 14 deletions(-) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index c78dc16..044da94 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -710,6 +710,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_tx_buf *tx_req; int rc; + unsigned long flags; if (unlikely(skb->len > tx->mtu)) { ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n", @@ -742,27 +743,36 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ skb_orphan(skb); skb_dst_drop(skb); + spin_lock_irqsave(&priv->lock, flags); + if (++priv->tx_outstanding == ipoib_sendq_size) { + ipoib_dbg(priv, "TX ring 0x%x full, stopping kernel net queue\n", + tx->qp->qp_num); + netif_stop_queue(dev); + } + spin_unlock_irqrestore(&priv->lock, flags); + if (netif_queue_stopped(dev)) { + rc = ib_req_notify_cq(priv->send_cq, + IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); + if (rc < 0) + ipoib_warn(priv, "request notify on send CQ failed\n"); + else if (rc) + ipoib_send_comp_handler(priv->send_cq, dev); + } + rc = post_send(priv, tx, tx->tx_head & (ipoib_sendq_size - 1), tx_req); if (unlikely(rc)) { ipoib_warn(priv, "post_send failed, error %d\n", rc); ++dev->stats.tx_errors; + spin_lock_irqsave(&priv->lock, flags); + --priv->tx_outstanding; + if (netif_queue_stopped(dev)) + netif_wake_queue(dev); + spin_unlock_irqrestore(&priv->lock, flags); ipoib_dma_unmap_tx(priv, tx_req); dev_kfree_skb_any(skb); } else { dev->trans_start = jiffies; ++tx->tx_head; - - if (++priv->tx_outstanding == ipoib_sendq_size) { - ipoib_dbg(priv, "TX ring 0x%x full, stopping kernel net queue\n", - tx->qp->qp_num); - netif_stop_queue(dev); - rc = ib_req_notify_cq(priv->send_cq, - IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); - if (rc < 0) - ipoib_warn(priv, "request notify on send CQ failed\n"); - else if (rc) - ipoib_send_comp_handler(priv->send_cq, dev); - } } } @@ -796,10 +806,13 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) netif_tx_lock(dev); ++tx->tx_tail; + + spin_lock_irqsave(&priv->lock, flags); if (unlikely(--priv->tx_outstanding == ipoib_sendq_size >> 1) && netif_queue_stopped(dev) && test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) netif_wake_queue(dev); + spin_unlock_irqrestore(&priv->lock, flags); if (wc->status != IB_WC_SUCCESS && wc->status != IB_WC_WR_FLUSH_ERR) { @@ -1169,6 +1182,7 @@ static void ipoib_cm_tx_destroy(struct ipoib_cm_tx *p) struct ipoib_dev_priv *priv = netdev_priv(p->dev); struct ipoib_tx_buf *tx_req; unsigned long begin; + unsigned long flags; ipoib_dbg(priv, "Destroy active connection 0x%x head 0x%x tail 0x%x\n", p->qp ? p->qp->qp_num : 0, p->tx_head, p->tx_tail); @@ -1198,10 +1212,12 @@ timeout: dev_kfree_skb_any(tx_req->skb); ++p->tx_tail; netif_tx_lock_bh(p->dev); + spin_lock_irqsave(&priv->lock, flags); if (unlikely(--priv->tx_outstanding == ipoib_sendq_size >> 1) && netif_queue_stopped(p->dev) && test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) netif_wake_queue(p->dev); + spin_unlock_irqrestore(&priv->lock, flags); netif_tx_unlock_bh(p->dev); } diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index d266667..7616e3c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -377,6 +377,7 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) struct ipoib_dev_priv *priv = netdev_priv(dev); unsigned int wr_id = wc->wr_id; struct ipoib_tx_buf *tx_req; + unsigned long flags; ipoib_dbg_data(priv, "send completion: id %d, status: %d\n", wr_id, wc->status); @@ -397,10 +398,13 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) dev_kfree_skb_any(tx_req->skb); ++priv->tx_tail; + + spin_lock_irqsave(&priv->lock, flags); if (unlikely(--priv->tx_outstanding == ipoib_sendq_size >> 1) && netif_queue_stopped(dev) && test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) netif_wake_queue(dev); + spin_unlock_irqrestore(&priv->lock, flags); if (wc->status != IB_WC_SUCCESS && wc->status != IB_WC_WR_FLUSH_ERR) { @@ -540,6 +544,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_tx_buf *tx_req; int hlen, rc; void *phead; + unsigned long flags; if (skb_is_gso(skb)) { hlen = skb_transport_offset(skb) + tcp_hdrlen(skb); @@ -587,12 +592,22 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, else priv->tx_wr.send_flags &= ~IB_SEND_IP_CSUM; + spin_lock_irqsave(&priv->lock, flags); if (++priv->tx_outstanding == ipoib_sendq_size) { ipoib_dbg(priv, "TX ring full, stopping kernel net queue\n"); if (ib_req_notify_cq(priv->send_cq, IB_CQ_NEXT_COMP)) ipoib_warn(priv, "request notify on send CQ failed\n"); netif_stop_queue(dev); } + spin_unlock_irqrestore(&priv->lock, flags); + if (netif_queue_stopped(dev)) { + rc = ib_req_notify_cq(priv->send_cq, + IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); + if (rc < 0) + ipoib_warn(priv, "request notify on send CQ failed\n"); + else if (rc) + ipoib_send_comp_handler(priv->send_cq, dev); + } skb_orphan(skb); skb_dst_drop(skb); @@ -602,11 +617,13 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, if (unlikely(rc)) { ipoib_warn(priv, "post_send failed, error %d\n", rc); ++dev->stats.tx_errors; + spin_lock_irqsave(&priv->lock, flags); --priv->tx_outstanding; - ipoib_dma_unmap_tx(priv, tx_req); - dev_kfree_skb_any(skb); if (netif_queue_stopped(dev)) netif_wake_queue(dev); + spin_unlock_irqrestore(&priv->lock, flags); + ipoib_dma_unmap_tx(priv, tx_req); + dev_kfree_skb_any(skb); } else { dev->trans_start = jiffies; @@ -825,6 +842,7 @@ int ipoib_ib_dev_stop(struct net_device *dev) unsigned long begin; struct ipoib_tx_buf *tx_req; int i; + unsigned long flags; if (test_and_clear_bit(IPOIB_FLAG_INITIALIZED, &priv->flags)) napi_disable(&priv->napi); @@ -857,7 +875,9 @@ int ipoib_ib_dev_stop(struct net_device *dev) ipoib_dma_unmap_tx(priv, tx_req); dev_kfree_skb_any(tx_req->skb); ++priv->tx_tail; + spin_lock_irqsave(&priv->lock, flags); --priv->tx_outstanding; + spin_unlock_irqrestore(&priv->lock, flags); } for (i = 0; i < ipoib_recvq_size; ++i) {